BOREAL
ENV.
RES. Vol. 7 • Diatom-inferred
eutrophication
and sedimentary phosphorus
27
BOREAL
ENVIRONMENT
RESEARCH 7:
27–40
ISSN 1239-6095
Helsinki 28 March 2002
© 2002
Diatom-inferred increase in limnetic
phosphorus concentration and the
associated changes in sedimentary
phosphorus fractions in Valkjärvi, a lake in
Kärkölä, Finland
Tommi Kauppila1)2), Teppo Moisio1)3) and Veli-Pekka Salonen1)*
1)
Department of Geology, University of Turku, FIN-20014 Turku, Finland
(*email: [email protected])
2)
present address: Geological Survey of Finland, P.O. Box 96, FIN-02151
Espoo, Finland (email: [email protected])
3)
present address: Golder Associates, Ruosilankuja 3E, 00390 Helsinki,
Finland (email: [email protected])
Kauppila, T., Moisio, T. & Salonen, V.-P. 2002. Diatom-inferred increase in limnetic phosphorus concentration and the associated changes in sedimentary
phosphorus fractions in Valkjärvi, a lake in Kärkölä, Finland. Boreal Env. Res.
7: 27–40. ISSN 1239-6095
A diatom-based total phosphorus (TP) model developed for southern Finland is used to
infer past epilimnetic TP concentrations in Valkjärvi, a lake located in Kärkölä, Finland;
and the results are compared with sedimentary phosphorus fractions. The diatom-based
models accurately predicted epilimnetic TP concentrations as indicated by monitoring
data available since 1979. Altogether, concentrations have increased from 15 µg P l–1
in the 1940s to the present level of ~65 µg P l–1. After an initial slow increase, the lake
became eutrophicated rapidly in the 1960s as a result of municipal wastewater from the
Kärkölä commune. These nutrient inputs resulted in increased sedimentary phosphorus
concentrations. The NaOH-extracted phosphorus fraction in particular increased in the
profundal sediments. The changes in sedimentary phosphorus nevertheless occurred
later than the increase in diatom-inferred limnetic TP concentration.
Introduction
Eutrophication of lakes is a well-known environmental problem in many parts of the world
and, after an initial debate over the nutrient
mainly responsible for the phenomenon, phosphorus is now recognized as the factor that most
frequently determines the level of increased pro-
28
duction in many freshwater ecosystems (OECD
1982). As the deterioration in water quality
associated with eutrophication is of economic
importance, extensive attempts have been made
to restore lakes to their original trophic state.
Since no long-term monitoring data are usually
available to define the background conditions
of these lakes, palaeolimnological methods have
proved invaluable for the study of their nutrient
history.
Microfossil-based inference models, particularly ones employing diatoms (Bacillariophyceae), have been widely used to estimate past
epilimnetic phosphorus concentrations in lakes
(Hall and Smol 1999). Several diatom-total phosphorus (TP) models have been developed for various geographic areas during the last ten years,
and their performance has been tested against
measured limnetic concentrations in down-core
applications, e.g. by Bennion et al. (1995), Marchetto and Bettinetti (1995), Hall et al. (1997),
Rippey et al. (1997), Lotter (1998), Alefs and
Müller (1999), and Kauppila et al. (2002), with
encouraging results. It has been shown that sedimentary diatom assemblages can be used to reliably reconstruct past lake water TP concentrations and certain other environmental variables.
It may partly be because of these recent advances
that the use of palaeolimnology is included in
the new EU Water Framework Directive (European Parliament and Council 2000) as one of
the tools to determine reference conditions for
different types of watercourse.
Another method used to trace eutrophication
is the study of sedimentary phosphorus concentrations. In contrast to diatom models, sediment
phosphorus cannot be used to reconstruct lake
water phosphorus concentrations because of a
number of complicating factors (Engstrom and
Wright 1984). For instance, not all phosphorus
in lake water is incorporated into the sediments,
and the fraction that is eventually sedimented
can change with time (Gibson et al. 1996). Furthermore, post-depositional mobility can lead
to erroneous interpretations (Carignan and Flett
1981, Boström 1984, Itkonen and Olander 1997).
Sediments nevertheless play an important role in
lake metabolism as both a sink and a source of
nutrients. Sedimentary phosphorus is an important phosphorus compartment in the lake system,
Kauppila et al.
•
BOREAL ENV. RES. Vol. 7
and therefore analyses of sedimentary phosphorus are essential in applied palaeolimnological
eutrophication studies.
In accordance with the importance of these
two methods, a number of papers have been
published that make use of both diatom-TP
inferences and sedimentary phosphorus analyses
(Anderson et al. 1993, Anderson 1994, Anderson and Rippey 1994, Bennion et al. 1995,
Brenner et al. 1995, Brenner et al. 1996, Rippey
and Anderson 1996, Hall et al. 1997, Rippey
et al. 1997, Chen and Wu 1999). In all of
these, total sedimentary phosphorus was analyzed. Additional information on phosphorus
binding to sediment can be gained, however,
by applying a chemical fractionation procedure
(Pettersson et al. 1988), and attempts have been
made to estimate the importance of such sedimentary phosphorus fractions as organic P, Cabound P, Fe and Al-bound P, labile P, exchangeable P, reductant-soluble P, and algal available P.
Accordingly, various extraction schemes for the
fractionation of sedimentary phosphorus have
been developed (Pettersson et al. 1988, Jáuregui
and García-Sánchez 1993).
This present study discusses changes in sedimentary phosphorus fractions, employing one
of the commonly used extraction schemes, and
their relation to the diatom-inferred increase
in limnetic total phosphorus concentration in
Valkjärvi, currently eutrophic boreal lake. Two
short cores were taken from the lake and analyzed for diatoms and sedimentary phosphorus
fractions to investigate the effect of coring location on these variables. Water chemistry data
available for the lake since 1979 were used
to check the observations, and an attempt was
made to date a lake sediment sequence using
soot particle stratigraphy and cyclicities in deposition detected by detailed grain-size analysis.
Sampling site
Valkjärvi is a lake located in Kärkölä, southern
Finland, at approximately 25°15´E, 60°52´N
(Fig. 1). It is currently eutrophic, with an average surface-water total phosphorus (TP) concentration of 65 µg P l–1 in the 1990s. The
lake has an area of 1.5 km2 and a catchment
BOREAL ENV. RES. Vol. 7
• Diatom-inferred eutrophication and sedimentary phosphorus
of 22.5 km2. It is a single-basin lake with a
maximum depth of 10.5 m. Surface sediments in
the catchment consist of glaciolacustrine clays,
some peat bogs and glaciofluvial deposits with
a few outcrops of Precambrian bedrock surrounded by till. Most of the clay area around
the lake is cultivated and there are both holiday
cottages and permanent houses close to the
shoreline. The brook of Pyhäoja runs into the
lake from Kärkölä (Fig. 1).
Material and methods
Valkjärvi was sampled in January 1997 with
a wedge-shaped crust freeze corer (Renberg
1981). The core KVALK 297 was taken from
the deepest part of the lake (9.5 m), and KVALK
397 from a location close to the inlet of the
brook Pyhäoja in a water depth of 1.7 m (Fig. 1).
This location was also cored with a Russiantype corer (Jowsey 1966) to examine the sedimentary structures. The cores were inspected
visually and described in the field according to
the method of Troels-Smith (1955) and using a
Munsell“ colour chart. The frozen cores were
then wrapped in metal foil and plastic, transported to the laboratory in a frozen state, and
stored at –18 °C.
To investigate the possible cyclical nature
of sedimentation at the brook mouth, the core
KVALK 397 was subsampled with a scalpel in
continuous 1 mm slices and analysed for the
grain-size distribution of inorganic matter. Prior
to the analysis, organic matter was digested with
H2O2 and aggregates were disintegrated by lowpower ultrasonication (2 h). The samples were
analyzed using the Coulter Electronics LS 200
laser particle sizer.
The mean, median, and modal grain-sizes of
approximately 400 grain-size distributions were
used in time-series analyses by autocorrelation,
partial autocorrelation and spectral analysis. The
analyses were performed after removal of any
linear trend in the data and subtraction of the
overall mean using a five-observation Hamming
window for spectral density estimates. All the
time series analyses were performed using STATISTICA 5.0 software.
The soot particle stratigraphy of both cores
29
!
!
ValkjŠrvi
"
Fig. 1. Map of Valkjärvi and its location in Fennoscandia. Grey dots on the large map = calibration
set lakes, black dots on the inset map = coring
sites.
was analyzed in continuous 1-cm slices. Soot
particles, or spherical carbonaceous particles
(SCP), originating from fossil fuel combustion
have a characteristic shape and can be identified
using a light microscope (Rose 1990). Their
distribution in a sediment core can be compared
with the known history of fossil fuel consumption for the region to date the sediment sequence.
Data on annual oil and coal consumption for
Finland were obtained from the State Technical
Research Centre (VTT). Only fuel types causing
SCP emissions were included in the statistics.
As SCP are almost entirely composed of
elemental carbon (Wik and Natkanski 1990),
they are chemically inert, and they can, therefore, be separated out for microscopy by removing all other sediment components by sequential
chemical attacks. The method used is a modification of that of Rose (1990) involving oxidation
with H2O2, acid digestion with HCl, removal of
silicates with HF and another HCl digestion to
remove any calcium fluoride crystals formed in
the HF step. A known amount of Lycopodium
spores was added to the sample after the oxidation step to facilitate quantitative counting of
30
the soot particles. The slides were mounted
with Naphrax‚ and studied under a Leitz Dialux
20EB microscope at 400¥ magnification. To
avoid bias in the counting procedure, all the
slides were recoded so that no indication of sediment depth was visible on them. All the counts
included at least 200 Lycopodium grains. The
SCP and grain-size data have been discussed
earlier by Kauppila et al. (2002), but the results
will be examined more thoroughly here based
on historical information about disturbances in
the lake.
Subsamples for the diatom analyses were
taken in 0.5 cm slices at 1.5 cm intervals and
treated with H2O2 to oxidize organic matter
and washed three times with distilled water to
remove the oxidant. A drop of the resulting
suspension was dried on a coverslip at room
temperature and the dried coverslip mounted
with Naphrax®. A minimum of 300 valves were
identified on each slide, mainly using Krammer
and Lange-Bertalot (1986–1991) as reference.
Author citations are included for taxa not named
according to this reference work. The microscope
used was a Leitz Orthoplan with a 100¥ Plan
Apo oil immersion phase objective (n.a. = 1.32)
and 12.5¥ eyepieces. The diatom results for
KVALK 297 have been published earlier in
Kauppila et al. (2002). These previous observations will be re-evaluated here by different
methods and compared with the results of core
KVALK 397.
The calibration data set used for the TP
reconstructions has been described by Kauppila
et al. (2002), who developed a weighted averaging partial least squares (WA-PLS) inference
model for the autumnal lake surface water TP
concentration. The 61-lake data set making up
their final model was used in the present study
to develop four separate WA-based models for TP
(inverse and classical deshrinking, with and without tolerance downweighting). The data consist
of surface sediment diatom counts and mostly
autumnal circulation period water chemistries
for lakes in the southern part of Finland (Fig. 1)
and have a TP range of 3–89 µg P l–1. Valkjärvi
lies, therefore, at the upper end of the calibration
set TP gradient. The models were developed
with the CALIBRATE 0.85 software of S. Juggins and C.J.F. ter Braak (1997, unpublished
Kauppila et al.
•
BOREAL ENV. RES. Vol. 7
software).
The suitability of the calibration dataset
for reconstruction of TP from fossil diatom
assemblages was assessed using constrained
ordinations and analogue matching performed
using the programs CANOCO version 4.0
(ter Braak & Òmilauer 1998) and MAT (S. Juggins, unpublished software), respectively. The
response of the assemblages to gradients other
than TP, indicated by a lack-of-fit of the fossil
samples with TP, was studied by entering them
as passive samples in Canonical Correspondence Analysis (CCA), constrained to TP. Samples with a squared residual distance from the
TP axis that was longer than the extreme 5%
of distances of the training set samples were
considered to have a poor fit to TP. Samples
having poor analogues in the calibration set
were defined as those having a squared chord
distance to the closest sample in the calibration
set that was larger than 0.44, corresponding to
the fifth percentile of the distances between
all training set samples. Samples with a minimum distance larger than 0.51 (tenth percentile)
were regarded as having very poor modern
analogues.
The phosphorus fractionation procedure,
modified from Hieltjes and Lijklema (1980),
involved sequential extractions with H2O, 0.1 M
NaOH and 0.5 M HCl and subsequent analysis
by the molybdenum blue method of Murphy
and Riley (1962). Separate subsamples were
taken for total phosphorus determination by the
molybdenum blue method after digestion of the
ignited samples with HNO3 and HCl (Bengtsson
and Enell 1986). In addition, the water content
of the samples was determined as weight loss
after drying overnight at 105 °C and loss-onignition after ashing the sample at 550 °C for
2 h.
The H2O extraction result is considered
to represent labile phosphorus (H2O-P), while
NaOH extracts phosphorus bound to short-range
ordered Fe and Al hydroxides and organic complexes that is exchangeable for OH– (NaOH-P).
This fraction is considered critical for internal
loading since the reduction of Fe(III) to Fe(II)
associated with seasonal anoxia in eutrophic
lakes often results in remobilization of phosphorus in sediment (Boström et al. 1982, Psen-
BOREAL ENV. RES. Vol. 7
• Diatom-inferred eutrophication and sedimentary phosphorus
ner 1984). In contrast, the HCl-extracted phosphorus (HCl-P), often called Ca-phosphorus, is
bound to carbonates and apatitic minerals and
is not liberated under reducing conditions. Due
to its largely mineral nature, this fraction is
frequently of importance in littoral sediments
(Boström et al. 1982). The residual fraction is
obtained as the difference between total phosphorus content and the sum of all fractions
and represents mainly phosphorus in refractory
organic material (Pettersson et al. 1988).
The municipality of Kärkölä has gathered
water monitoring data for Valkjärvi, largely
through the agency of environmental consulting
companies that have been responsible for taking
the samples and analyzing them over the years.
The determinations have been carried out according to accepted standards for the analysis of
fresh water. In practice, total phosphorus analysis involved peroxodisulfate digestion and photometric determination using the molybdenum
blue method (e.g., Suomen standardisoimisliitto
SFS ry. 1986)
KVALK 297
Da
Da
nig.4, strf.2, elas.0, sicc.2
colour: 2,5Y 2,5/1
struc. gyttja with irregular
sulphide layers
comp. Dg2, Ag1, As1, sulf+
28 cm
Db
Db
77 cm
KVALK 397
5 cm
Da
Da
Db
29 cm
Db
Results and discussion
Lithology
Dc
Dc
The core taken from the deepest part of the
lake (KVALK 297), 77 cm in length, consists
of homogeneous silty gyttja up to the 28 cm
level (Fig. 2). The sediment section is dark
olive green in colour, as is typical for lakes
in southern Finland with high autochthonous
production and a clay overburden in their catchment. Dark sulphide laminae appear in the sediment at approximately 28 cm, but the structure
is too uneven to allow any estimates of sedimentation rate to be made even if the stripes were
formed annually. The sulphide-coloured section
continues up to the top of the sediment, indicating that reducing conditions have occurred
intermittently in the upper sediment layer for
several years now.
At the littoral KVALK 397 site, the additional
sampling with a Russian-type peat corer allowed
penetration of the lake sediment sequence down
to the underlying laminated silty clay. The gyttja
resembles that of the KVALK 297 core but is
31
110 cm
113 cm
Dd
Dd
nig.3, strf.0, elas.0, sicc.2
colour: 10YR 3/1
struc. homogeneous silty
gyttja
comp. Dg2, Ag1, As1
nig. 4, strf. 0, elas. 0, sicc. 2
colour: 2,5Y 3/1 and 2,5Y 2,5/1
struc. loose silty gyttja
comp. Dg1,Ld1,Ag2,As+
nig. 2,5, strf. 2, elas.0, sicc.2
colour: 2,5Y 3/1
struc. silty, unevenly laminated
gyttja
comp. Ag2, Dg1, Ld1, As+, sulf+
nig.3, strf.0,5,elas.0,sicc.2
colour: 2,5Y 3/1
struc. homogeneous
gyttja
comp. Ag1,5, Dg2, Ld0,5,As+
nig.2, strf.1,elas.1,sicc.2
colour: struc. laminated silty clay
comp. As2, Ag1,5, Ld0,5
Fig. 2. Core lithologies according to the Troels-Smith
system. Da-Dd = loci descripti. The graphical symbols
do not conform to Troels-Smith (1955).
slightly richer in silt and is thus lighter in colour.
Despite the different sedimentary environment
and the slight difference in the appearance of
the gyttja, sulphide stripes become visible in
KVALK 397 at approximately the same level as
in the deep-water core (29 cm), but the loose
top section found in KVALK 397 is lacking in
KVALK 297.
Kauppila et al.
32
•
BOREAL ENV. RES. Vol. 7
Fig. 4. Grain-size analysis results (median) for core
KVALK 397. Dashed lines mark the section boundaries
used in time series analysis.
Fig. 3. Vertical distribution of water content and
percentage loss-on-ignition (LOI) in the cores.
Black = KVALK 297, grey = KVALK 397.
Loss-on-ignition and water content
Loss-on-ignition (LOI) and water content (W%)
profiles for the sediment section between 0 and
40 cm are presented for both cores in Fig. 3.
There is a section of increased LOI and W%
between 36 and 27 cm in KVALK 297, whereas
LOI declines rapidly between 28 and 32 cm in
KVALK 397 and the water content decreases
immediately from 40 cm. There is a section of
low LOI and water content above 27 cm in
both cores, with a subsequent increase towards
the sediment surface, probably due to increased
productivity and to ongoing compaction and
organic matter mineralization.
Grain size and soot particles
The results of the grain-size analyses are presented in Fig. 4, which displays the variation in
median grain size (µm) in core KVALK 397 for
the top 40 cm. The mean and modal grain size
(not shown) produce similar results. The profile
can be divided into three sections with differing
characteristics. The first section ranges from 40
cm to 28 cm and is characterized by a regular
fluctuation in median grain size around 8.5 µm.
The amplitude of these cycles increases upwards
and an abrupt decrease in amplitude marks the
beginning of the second section at 28 cm. In this
section, extending to 19 cm, the average median
grain size is approximately similar to that of the
lower section but the variations are less regular
and the size range is smaller. In contrast, the
uppermost section is characterized by coarser
material and large variation. The transition at 19
cm is again an abrupt one.
The estimation of deposition rates was based
on cyclicities in the grain size of the deposited
inorganic matter. As expected from examination
of the grain-size curve (Fig. 4), a cyclicity was
identifiable for the deepest sediment section by
all three time series analysis methods employed.
Assuming that one cycle corresponds to one
year’s deposition, all the methods suggested a
deposition rate of 10 mm a–1 for this seemingly
structureless section between 40 and 28 cm. In
contrast, there was no agreement between the
methods for the mid-section. A deposition rate
of 7 mm a–1 will be used for the 28–19 cm
section in the following discussion, based on
partial autocorrelations and, to a lesser extent,
autocorrelations. Similarly, the deposition rate
of 8.5 mm a–1 inferred for the 19–0 cm section
is based on the somewhat differing cyclicities
detected by the three methods. Lags of 8 and
9 mm had the highest partial autocorrelations for
the median and modal grain sizes, the highest
autocorrelations being associated with lags of
9 and 10 mm, and spectral analysis suggested
a period of 9.5 mm for this topmost section.
The selected estimate is thus a consensus figure
based on all three methods.
Soot particle dating was used to support the
BOREAL ENV. RES. Vol. 7
• Diatom-inferred eutrophication and sedimentary phosphorus
33
&&
&&
&
&
&'
&'
&
"#$%" &'
"#$%" &'
( )
! &
&
&
Fig. 5. Soot particle profiles for cores KVALK 297 and
KVALK 397. Yearly consumption of SCP-producing
forms of oil and coal in Finland is expressed as energy
(PJ = petajoule). The year scale is based on the
estimated deposition rates. Modified from Kauppila
et al. (2002).
grain-size inferred deposition rates. The results
of SCP analysis for both cores are presented in
Fig. 5. The profiles are broadly similar, with
soot particles appearing at 31 cm, after which
the concentration rapidly increases. The most
marked increase in fossil fuel consumption in
Finland occurred in the 1960s, and can be detectable with SCP to date the sediments (see Tolonen et al. 1990). The SCP concentrations are
higher in KVALK 297 and the curve less noisy,
presumably because of the funnelling effect that
deposits more soot at the deepest point of the
lake. The KVALK 297 profile is therefore easier
to compare with the statistics of annual oil and
coal consumption in Finland (Fig. 5). In view of
the similar lithologies of the two cores and
the appearance of SCP at the same level, the
fossil fuel consumption curve has been fitted to
both SCP profiles using the grain size-inferred
Fig. 6. Median and kurtosis for the grain-size distributions in samples between 30 and 40 cm depth in
KVALK 397, showing the change at 37 cm.
deposition rates for KVALK 397. The resulting
fit is good, lending support to the estimated
deposition rates.
The use of detailed grain size analyses thus
appears to facilitate deposition rate estimates
under favourable conditions. The inferred deposition rates for KVALK 397 are not unambiguous, however, and neither of the prominent
changes in grain size can be dated to 1952, a
year with known disturbances (dredging of the
Pyhäoja brook and the beginning of lake level
regulation). Instead, the 1952 sediment level
falls at 37 cm, just below the changes in LOI and
water content. These changes could have resulted
in more minerogenic matter being deposited at
the brook mouth and a pulse of organic-rich sediment being redeposited from near-shore areas
into the deeper parts as a consequence of water
level regulation. Furthermore, the median grain
size increases slightly at the same level, the
cyclicity in kurtosis disappears, and the grain
size distributions become less peaked overall
(lower kurtosis, see Fig. 6). These changes may
also be related to the 1952 disturbances, but
the shift in modal grain size observable at 19
cm remains unexplained. The irregularity in the
grain size curve that begins at 28 cm, on the
other hand, may be related to sulphide grains,
as sulphide colouring appears in the sediment
at this level.
•
Kauppila et al.
BOREAL ENV. RES. Vol. 7
# %
" # $ 34
&
#
'
#
(
*
+ #
"
#
!
! )
'
Fig. 7. Distribution of
selected diatom taxa in
the profundal KVALK 297
core. Solid line = 5¥ exaggeration. Plankton percentage is also shown.
Modified from Kauppila et
al. (2002).
%
*
#
)
#
+
$
#
&
' #
Diatoms
The relative abundances of selected diatom taxa
are shown in Figs. 7 and 8 (KVALK 297 and
!
(
)
Fig. 8. Relative abundances of selected diatom
taxa in the littoral KVALK
397 core. Plankton percentage is also shown.
Solid line = 5¥ exaggeration.
397, respectively). The dendrograms, produced
with CONISS software, which uses the incremental sum of squares method for stratigraphically constrained cluster analysis (Grimm 1987),
Table 1. Jack-knifed estimated model parameters for the four models used in the reconstruction. RMSEP = root
mean squared error of prediction.
——————————————————————————————————————————————————————————————————
R2
Max. bias (log µg P l–1)
Deshrinking
Model
RMSEP (log µg P l–1)
——————————————————————————————————————————————————————————————————
Inverse
WA
0.1604
0.7586
0.2747
Inverse
WA tol
0.1691
0.7350
0.2939
Classical
WA
0.1673
0.7632
0.1309
Classical
WA tol
0.1704
0.7411
0.1555
——————————————————————————————————————————————————————————————————
BOREAL ENV. RES. Vol. 7
• Diatom-inferred eutrophication and sedimentary phosphorus
are also shown. The 40 most abundant taxa
were used for clustering, and local diatom zones
(L.D.Z.) were determined on the basis of the
analyst’s interpretation of the diatom distributions and the dendrogram.
The major species shifts are the same in both
diagrams. The lower section of both cores is
characterized by high abundances of Tabellaria
flocculosa, Aulacoseira distans and Cyclotella
radiosa, whereas Aulacoseira ambigua, Aulacoseira granulata v. angustissima, Stephanodiscus
hantzschii and Diatoma tenuis predominate in
the Kar 1 zone. The proportions of major taxa
also increase at corresponding levels in both
cores. Despite these similarities, the differences
caused by contrasting coring locations (littoral
vs. pelagic/profundal) are manifested in terms of
a higher relative abundance of planktonic taxa
in KVALK 297 and abundant benthic Fragilaria
in KVALK 397. Aulacoseira subarctica f. recta,
for instance, increases in zone Kar 1 in KVALK
297 but not in the littoral core.
The dominance of eutrophic planktonic taxa
in the Kar 1 zone results in a decline in the
relative abundances of periphytic Achnanthes,
Navicula and Fragilaria taxa that are present
throughout the cores. Logically, the effect will
be stronger in the profundal KVALK 297 core.
The taxon named here Synedra rumpens var.
familiaris (Kütz.) Grunow is present only in the
upper parts of the diagrams. It increases to its
maximum relative abundance at the top of zone
Kar 2 in both cores, but declines to very low
levels in the Kar 1 zone.
Inference model and calibration
The data set used for developing the inference
models is described in more detail by Kauppila
et al. (2002), who show by means of constrained
and partially constrained ordinations that it is
possible to reconstruct TP from fossil assemblages using the available data. The inference
models developed here are WA models with
and without tolerance downweighting, using both
classical and inverse deshrinking (classWA, classWAtol, invWA, invWAtol, Table 1) with logtransformed TP concentrations. As expected, the
inverse deshrinking models have lower root
35
Fig. 9. Diatom-inferred epilimnetic phosphorus concentration profiles for the KVALK 297 and KVALK 397
cores. Results obtained with all four models (weighted
averaging using classical and inverse deshrinking,
with and without tolerance downweighting) and their
means are shown.
mean squared errors of prediction (RMSEP)
than the models with classical deshrinking (ter
Braak and Juggins 1993) but have substantially
larger maximum bias.
The calibration results are shown in Fig. 9,
with curves to represent the means of the four
diatom-inferred TP values. In KVALK 297, the
reconstructed TP concentration remains at 15 µg
P l–1 between 56 and 44 cm, increasing gradually
to 25 µg P l–1 at the 25 cm level. Thereafter,
DI-TP increases rapidly until the values stabilize
around 65 µg P l–1 at 8 cm. In contrast, the
increase in DI-TP is less marked in KVALK
397 and no levelling out of the trend is seen at
the top. The DI-TP concentrations in the upper
section are 10 µg P l–1 lower for KVALK 397
than for KVALK 297.
The lower predicted TP concentration in the
littoral core is due to the higher number of
periphytic taxa in the samples. Due to their
high limnetic nutrient concentration, eutrophic
lakes have a high planktonic diatom abundance,
which causes a corresponding lowering in the
relative proportion of benthic diatoms in surface
sediment samples. The weighted averaging-based
regression methods may, therefore, assign too
low TP optima for the periphytic taxa. However, this need not always be the case as nonplanktonic taxa of especially the genus Fragi-
36
Fig. 10. Diatom-inferred TP profile for the KVALK 297
core. The measured epilimnetic TP concentrations
are placed in the figure using the estimated deposition
rates. Dashed lines = mean ‘error’ of prediction (see
text), arrows = late winter TP measurements in 1984
and 1986.
laria have in some sets been observed to occur
along the whole length of the TP gradient (Bennion 1995, Bennion et al. 2001). Therefore, such
taxa will be assigned TP optima close to the
centre of the gradient available and, depending
on the characteristics of the calibration set and
the target lake, this can lead to TP overestimation (Bennion et al. 2001).
There are also differences between the cores
in the predictions given with the four models.
Marked differences in DI-TP are obtained using
classical and inverse deshrinking above ~30 µg
P l–1 in both cores, because these types of model
have differing trends in their residuals.
Assessed by CCA constrained solely to TP,
only one sample, at 20 cm in KVALK 397,
exhibited a poor fit to TP as it had a residual distance from the TP axis longer than the extreme
5% of the distances among the training set samples. However, some samples lacked modern
analogues in the training set. In Fig. 9, samples
with poor modern analogues are marked with a
single dot and samples with very poor modern
analogues with two dots.
The modern analogue data also highlight the
differences between the cores, as most of the
KVALK 397 samples lack modern analogues.
This presumably results from the littoral coring
location, as the training set represents cores
from central lake locations. The KVALK 297
coring site is therefore more appropriate than
Kauppila et al.
•
BOREAL ENV. RES. Vol. 7
KVALK 397 for calibration purposes using the
present data set. The reason for the concentration of poor modern analogues in the section
immediately above 25 cm in both profiles apparently lies in the occurrence of Synedra rumpens
var. familiaris (Kütz.) Grunow at higher percentage abundance than in any calibration set
lake. The section is thicker in KVALK 297,
as is also evident from the diatom stratigraphy
(Fig. 8). Similarly, the lack of modern analogues
at the base of the KVALK 297 core probably
results from the high proportion of Tabellaria
flocculosa.
The diatom-inferred TP for the KVALK
297 core, expressed as the mean of the values
obtained using the four calibration methods,
was compared with the available epilimnetic TP
measurements (Fig. 10). The prediction error
shown in Fig. 10 (12.6 µg P l–1) was calculated
as the mean error of prediction using the mean
for the four jack-knifed inference models and
back transformed log TP concentrations. This
error is a considerable one, but much smaller
than the annual range of epilimnetic TP concentrations in Valkjärvi (45 µg P l–1 in 1994).
Understandably, the diatom-based model does
not predict the very high concentrations, such as
those measured in April 1984 and 1986 (shown
by arrows in Fig. 10), but DI-TP does follow
the autumnal measured TP values closely (diamonds in Fig. 10). This is to be expected,
as the model was developed using autumnal
circulation TP concentrations. The increase in
lake surface water TP between 1979 and 1987
and the subsequent stabilization are faithfully
recorded in the DI-TP results. The close match
between measured and inferred TP also indirectly supports the deposition rate estimates
discussed earlier. In addition, the mean of the
four WA models is a more accurate predictor
of the measured TP than the WA-PLS 1 component model described by Kauppila et al. (2002)
which here corresponds to the invWA model.
The results indicate that Valkjärvi has not
always been eutrophic. The increase in nutrient
concentrations started in the 1940s and intensified in the mid-1960s. Agriculture was probably responsible for the initial, slow increase
in DI-TP, whereas the rapid change in the mid1960s was caused by municipal wastewater from
• Diatom-inferred eutrophication and sedimentary phosphorus
Kärkölä. A sewer system was built in Kärkölä in
1961, conducting untreated wastewater into the
Pyhäoja brook. The rapid eutrophication phase
was probably intensified by seasonal anoxia at
or slightly below the sediment-water interface,
as evidenced by the appearance of sulphide
laminations in the sediment. From 1975 onwards,
however, the wastewater was directed to a treatment plant located downstream of Valkjärvi, and
this may have caused the temporary drop in
diatom-inferred TP at 18 cm, although it did not
reverse the increase in lake water phosphorus
concentration. This probably reflects the importance of diffuse and internal phosphorus loading
sources for lake eutrophication.
However, the diatom-inferred TP stabilizes
above 10 cm despite some changes in the proportions of major taxa, since all these taxa have
similar TP optima. Due to the large scatter,
no levelling out of nutrient concentrations after
1987 is evident from the monitoring data, when
taken as a whole, but the autumnal measured
TP concentrations do not increase after this
period and extremely high nutrient concentrations (> 100 µg P l–1) have not been measured
since 1986. The reason for this pattern cannot lie
in restoration efforts, which were started only
in 1993, although the proportions of Aulacoseira granulata var. angustissima and Stephanodiscus hantzschii, both taxa with high TP optima
(46.2 µg P l–1 and 35.5 µg P l–1, respectively),
decrease slightly in the topmost samples.
Sedimentary phosphorus fractions
The sedimentary phosphorus fractionation results
are presented in Fig. 11. The total phosphorus concentrations are comparable and increase upwards
in both cores, whereas the H2O-leached fraction
is low in all the samples. There is a rapid increase
in NaOH-P at 15 cm in KVALK 297, with
a simultaneous but smaller increase in HCl-P,
whereas both HCl and NaOH-extracted phosphorus increase above 15 cm in KVALK 397, with
the HCl fraction dominant in all the samples.
This is to be expected, as littoral sediments
often contain relatively much Ca-bound phosphorus (Boström et al. 1982). The section of
increased total phosphorus between 37 and 29 cm
"#$%" &'
37
"#$%" &'
BOREAL ENV. RES. Vol. 7
!
Fig. 11. Vertical distribution of sedimentary phosphorus
fractions in the Valkjärvi cores.
in KVALK 297 results from the elevated organic
matter content (Fig. 3), presumably caused by the
lake level regulation. The change is observable
in the NaOH fraction (Fe-bound P) since much
of this phosphorus is bound to Fe(III)-organic
matter complexes (Pettersson et al. 1988).
Sedimentary phosphorus concentrations begin
to increase later than the diatom-inferred TP
rise. The slow initial increase in epilimnetic
phosphorus concentration, as inferred from the
diatom data, from 40 cm sediment depth, did
not cause any increase in sediment phosphorus
content or any change in the relative proportions of the phosphorus fractions. It is not until
15 cm that the total phosphorus concentration and
the proportion of labile NaOH-P both increase,
approximately ten years after the diatom-inferred
limnetic phosphorus concentrations began to
increase more rapidly. There appears, therefore,
to be a delay between nutrient enrichment
and the increase in sedimentary phosphorus.
Aquatic biota are able to assimilate and recycle
an increased amount of nutrients in the water
column, thereby delaying any increase in phosphorus sedimentation, but the sedimentation of
phosphorus inevitably increases as limnetic concentrations rise further.
The sedimentary total phosphorus concentration between 3 and 13 cm in KVALK 397 does
not increase together with HCl-P and NaOH-P
as it does in KVALK 297, and the total phosphorus concentration is less than the sum of
38
the fractions in four samples (Fig. 11). Excluding the possibility of error in the analysis, this
phenomenon may be related to a change in the
composition of sedimentary phosphorus that has
caused a decrease in the residual phosphorus
fraction and lessened the efficacy of the acid
leaching used in total phosphorus determination.
In any case, it renders the total phosphorus
determination for the upper section of KVALK
397 unreliable.
The increase in total phosphorus and NaOH-P
in KVALK 297 is interrupted by a depression at
~5 cm not observable in the HCl fraction. This
is probably caused by post-depositional mobility
of phosphorus in the sediment, as the changes
are greatest in the NaOH fraction. It is this
fraction that becomes mobilized with a lowering of the redox potential in the sediment and
a reduction of Fe(III) to Fe(II) (Boström et
al. 1982). The topmost sediment still binds
phosphorus, however.
Conclusions
The results show that high-resolution grain-size
analyses can be used for approximate lake sediment dating even when the sediment looks homogeneous. In the present study the proximity of the
coring location to an inlet and a high deposition
rate contribute to the success of the approach.
Diatom-inferred total phosphorus concentrations followed the measured concentrations
closely and indicated that nutrient levels in
Valkjärvi have increased considerably since 1940.
In fact, the inferred TP levels span almost the
whole TP range of the lakes in the calibration
set. The initial increase was slow and probably
related to intensified agriculture after the war,
whereas municipal wastewater caused more rapid
eutrophication in the 1960s. The limnetic TP
concentration continued to increase even after
diversion of the wastewater, however, because
of internal lake processes and/or loading from
other sources.
The diatom-based inferences for the littoral
core differ from those for the pelagic core,
with lower inferred TP concentrations and fewer
modern analogues for the former. Both features
result from the fact that cores were taken from
Kauppila et al.
•
BOREAL ENV. RES. Vol. 7
central lake locations in the training set. Thus
littoral diatoms are poorly represented in the
surface sediments of the nutrient-rich calibration
set lakes, leading to their being assigned low
estimated TP optima. It is perhaps unsurprising
that a calibration set based on sediment samples
from the central area of lakes will not provide
good analogues for diatom assemblages found
in littoral cores.
As expected, sedimentary phosphorus concentrations increase with the phosphorus concentration in the lake water, but there is a delay
in the response, as small initial increases in
limnetic [TP] are not reflected in sedimentary
phosphorus. Nutrient recycling appears to delay
the increase in phosphorus sedimentation, but
once the increase does happen, the fractional
composition of the sediment phosphorus also
changes, especially in the profundal core. Most
of the increase is in NaOH-extracted phosphorus, the fraction susceptible to remobilization
under reducing conditions. Indeed, there are
signs of post-depositional mobility in the uppermost sediment layers in the profundal core,
which makes interpretation of the sedimentary
phosphorus profiles difficult.
Finally, the results highlight the importance
of combining several approaches in sedimentbased eutrophication studies. Explanations for
observed patterns are more likely to be found
using multiple analyses, and a well-selected
package of analyses can convey a wealth of
information needed for lake restoration purposes
and for evaluating the current status of lakes.
Acknowledgements: The authors wish to extend their
thanks to the Maj and Tor Nessling Foundation for its
financial support for the first phase of the project. The
diatom-based calibrations were made possible by a grant
to TK from the Emil Aaltonen Foundation. We also wish
to thank all those who helped with the fieldwork, and also
Riitta Hyytiäinen and Jaakko Airola in Kärkölä for all
their practical help and discussions. Malcolm Hicks and
Emily Bradshaw reviewed the English of the manuscript.
References
Alefs J. & Müller J. 1999. Differences in the eutrophication dynamics of Ammersee and Starnberger See
(Southern Germany), reflected by the diatom succession in varve-dated sediments. J. Paleolim. 21:
395–407.
BOREAL ENV. RES. Vol. 7
• Diatom-inferred eutrophication and sedimentary phosphorus
Anderson N.J. 1994. Comparative planktonic diatom biomass responses to lake and catchment disturbance. J.
Plankton Res. 16: 133–150.
Anderson N.J. & Rippey B. 1994. Monitoring lake recovery from point-source eutrophication: the use of diatom-inferred epilimnetic total phosphorus and sediment chemistry. Freshwat. Biol. 32: 625–639.
Anderson N.J., Rippey, B. & Gibson C.E. 1993. A comparison of sedimentary and diatom-inferred phosphorus profiles: implications for defining pre-disturbance
nutrient conditions. Hydrobiologia 253: 357–366.
Bengtsson L. & Enell M. 1986. Chemical analysis. In:
Berglund B. (ed.), Handbook of Holocene palaeoecology and palaeohydrology, John Wiley & Sons,
London, pp. 433–436.
Bennion H. 1995. Surface sediment diatom assemblages
in shallow, artificial, enriched ponds, and implications
for reconstructing trophic status. Diat. Res. 10: 1–19.
Bennion H., Appleby P.G. & Phillips G.L. 2001.
Reconstructing nutrient histories in the Norfolk
Broads, UK: implications for the role of diatom-total
phosphorus transfer functions in shallow lake
management. J. Paleolim. 26: 181–204.
Bennion H., Wunsam S. & Schmidt R. 1995. The validation of diatom-phosphorus transfer functions: an
example of Mondsee, Austria. Freshwat. Biol. 34:
271–283.
Boström B. 1984. Potential mobility of phosphorus in
different types of lake sediment. Int. Revue ges.
Hydrobiol. 69: 457–474.
Boström B., Jansson, M. & Forsberg, C. 1982. Phosphorus release from lake sediments. In: Elster H.-J. &
Ohle W. (eds), Nutrient remobilization from sediments and its limnological effects. Arch. Hydrobiol.
Beih. Ergebn. Limnol. 18: 5–60.
Brenner M., Whitmore T.J., Curtis J. & Shelske C.L.
1995. Historical ecology of a hypereutrophic Florida
lake. Lake Reserv. Manage. 11: 255–271.
Brenner M., Whitmore T.J. & Schelske C.L. 1996. Paleolimnological evaluation of historical trophic state
conditions in hypereutrophic Lake Thonotosassa, Florida, USA. Hydrobiologia 331: 143–152.
Carignan R. & Flett R.J. 1981. Postdepositional mobility
of phosphorus in lake sediments. Limnol. Oceanogr.
26: 361–366.
Chen S.-H. & Wu J.-T. 1999. Paleolimnological environment indicated by the diatom and pollen assemblages
in an alpine lake in Taiwan. J. Paleolim. 22: 149–158.
Engstrom D.R. & Wright H.E.Jr. 1984. Chemical stratigraphy of lake sediments as a record of environmental
change. In: Haworth E.Y. & Lund J.W.G. (eds), Lake
sediments and environmental history. Studies in palaeolimnology and palaeoecology in honour of Winifred
Tutin, Leicester University Press, Leicester, pp. 11–67.
European Parliament and Council 2000. Directive
2000/60/EC of the European Parliament and of the
Council of 23 October 2000 establishing a framework
for Community action in the field of water policy.
Official Journal L 327: 1–71.
39
Gibson C.E., Foy R.H. & Bailey-Watts A.E. 1996. An
analysis of the total phosphorus cycle in some temperate lakes: the response to enrichment. Freshwat.
Biol. 35: 525–532.
Grimm E.C. 1987. CONISS: a Fortran 77 program for
stratigraphically constrained cluster analysis by the
method of incremental sum of squares. Comp. Geosci.
13: 13–35.
Hall R.I. & Smol J.P. 1999. Diatoms as indicators of lake
eutrophication. In: Stoermer E.F. & Smol J.P. (eds.),
Diatoms: applications for the environmental and earth
sciences. Cambridge University Press, Cambridge,
pp. 128–168.
Hall R.I., Leavitt P.R., Smol J.P. & Zirnhelt N. 1997.
Comparison of diatoms, fossil pigments and historical
records as measures of lake eutrophication. Freshwat.
Biol. 38: 401–417.
Hieltjes A.H.M. & Lijklema L. 1980. Fractionation of
inorganic phosphates in different types of lake sediment. J. Environ. Qual. 9: 405–407.
Itkonen A. & Olander H. 1997. The origin of the hypertrophic state of a shallow boreal shield lake. Boreal
Env. Res. 2: 183–198.
Jáuregui J. & García Sánchez J.A. 1993. Fractionation of
sedimentary phosphorus: A comparison of four methods. Verh. Internat. Verein. Limnol. 25: 1150–1152.
Jowsey P.C. 1966. An improved peat sampler. New Phytol.
65: 245–248.
Kauppila T., Moisio T. & Salonen V.-P. 2002. A diatombased inference model for autumn epilimnetic total
phosphorus concentration and its application to a presently eutrophic boreal lake. J. Paleolim. 27. [In press].
Krammer K. & Lange-Bertalot H. 1986–1991: Bacillariophyceae 1–4. In: Ettl H., Gärtner G., Gerloff J.,
Heynig H. & Mollenhauer D. (eds.), Süsswasserflora
von Mitteleuropa, Band 2. Gustav Fisher, Jena, Stuttgart, Lübeck, Ulm.
Lotter A.F. 1998. The recent eutrophication of Baldeggersee (Switzerland) as assessed by fossil diatom
assemblages. The Holocene 8: 395–405.
Marchetto A. & Bettinetti R. 1995. Reconstruction of
the phosphorus history of two deep, subalpine Italian
lakes from sedimentary diatoms, compared with longterm chemical measurements. Mem. Ist. ital. Idrobiol.
53: 27–38.
Murphy J. & Riley J.P. 1962. A modified single solution
method for the determination of phosphate in natural
waters. Anal. Chim. Acta 27: 31–36.
OECD, 1982. Eutrophication of waters. Monitoring,
assessment and control. Organisation for economic
co-operation and development, Paris, 154 pp.
Pettersson K., Boström B. & Jacobsen O.-S. 1988. Phosphorus in sediments — speciation and analysis. Hydrobiologia 170: 91–101.
Psenner R. 1984. Phosphorus release patterns from sediments of a meromictic mesotrophic lake (Piburger
See, Austria). Verh. Int. Ver. Limnol. 22: 219–228.
Renberg I. 1981. Formation, structure and visual appearance of iron-rich varved lake sediments. Verh. Int.
40
Kauppila et al.
Ver. Limnol. 21: 94–101.
Rippey B. & Anderson N.J. 1996. Reconstruction of lake
phosphorus loading and dynamics using the sedimentary record. Environ. Sci. Technol. 30: 1786–1788.
Rippey B., Anderson N.J. & Foy R.H. 1997. Accuracy
of diatom-inferred total phosphorus concentrations
and the accelerated eutrophication of a lake due to
reduced flushing and increased internal loading. Can.
J. Fish. Aquat. Sci 54: 2637–2646.
Rose N.L. 1990. A method for the extraction of carbonaceous particles from lake sediment. J. Paleolim. 3:
45–53.
Suomen standardisoimisliitto SFS ry. 1986. Veden
kokonaisfosforin määritys. Hajotus peroksodisulfaatilla. SFS 3026:1986, 11 pp.
ter Braak C.J.F. & Juggins S. 1993. Weighted averaging partial least squares regression (WA-PLS): an improved
method for reconstructing environmental variables
•
BOREAL ENV. RES. Vol. 7
from species assemblages. Hydrobiologia 269/270:
485–502.
ter Braak C.J.F. & Òmilauer P. 1998. CANOCO reference
manual. Centre for Biometry Wageningen, Wageningen, 351 pp.
Tolonen K., Haapalahti R. & Suksi J. 1990. Comparison
of varve dated soot ball chronology and lead-210
dating in Finland. In: Saarnisto M. & Kahra A. (eds.),
Laminated sediments. Proceedings of the workshop at
Lammi biological station, 4–6 June 1990. Geological
Survey of Finland, Special Paper 14: 65–75.
Troels-Smith J. 1955. Karakterisering af lose jordarter —
Characterization of unconsolidated sediments. Geological Survey of Denmark, Series IV 3: 10, 73 pp.
Wik M. & Natkanski J. 1990. British and Scandinavian
lake sediment records of carbonaceous particles from
fossil-fuel combustion. Phil. Trans. R. Soc. Lond. B
327: 319–323.
Received 28 June 2001, accepted 29 October 2001